11
On the Development of Lexical and Non-Lexical
Spelling Procedures of French-Speaking
Normal and Disabled Children
JESUS ALEGRIA AND PHILIPPE MOUSTY
Université libre de Bruxelles
The French orthography, despite some peculiarities, belongs to the family of deep
orthographic systems (Gak, 1976; Catach, 1980), as opposed to shallow or superficial
ones. Shallow systems represent language at the phonological level. Deep systems
partially represent language at the phonological level too, but simultaneously include
aspects of morphology and syntax (Klima, 1972; Gleitman & Rozin, 1977). This often
creates conflicting situations. For example, to represent the morphology of French
verbs, unpronounced letters are added. As a result, different spellings are associated
to some homophonous members of the same morphosyntactic paradigm:
il change/fô/ja/
ils changent/J"Q33/
(he changes)
(they change)
changer/Ja3e/
changé/J"fl3e/
er infinitive (to change)
é past participle (changed)
These examples show that morphological and syntactic aspects of language
that are not marked at the phonological level are, nevertheless, orthographically
represented. French orthography also presents complexities at the phonological level,
as shown in the following examples:
Handbook of Spelling: Theory, Process and Intervention. Edited by G. D. A. Brown and N. C. Ellis.
©1994 John Wiley & Sons Ltd.
212
ABNORMAL SPELLING
/ / / -» ch is consistent
/â/->fl« is inconsistent, en is a frequent alternative
/ 3 / ^ g is inconsistent, j is a plausible alternative
In the first case, phonological knowledge is sufficient to spell correctly the///
phoneme. In the two others, lexical knowledge is necessary: change contains an (not
en) and g (not y')- Furthermore, in both examples of inconsistency, the intra-word
context comes into play to constrain spelling:
/ a / -» an is more frequent than / a / -> en within words
/ a / -» e« is more frequent than / a / -> o« at the beginning of words
/3/^i? or y* before e and /. Before a, o and u, the letter g cannot be used
Véronis (1988) conducted a computer simulation study aimed at evaluating the
productivity of sound-to-spelling transcription rules (SSTRs). The corpus contained
3724 current words (function words were excluded) including 19072 phonemes. The
results show that ab6ut 50 SSTRs allow 88% of the phonemes to be transcribed'
correctly but only about 50% of the words. An increase in the number of rules did
not improve the spelling performance of the model. So an SSTR-based strategy
permits not more than half of the French words to be spelled correctly. The other
words contain spelling irregularities that cannot be derived from the phonology, It
is clear that in order to spell French correctly, the child must possess and use a number
of linguistic abilities that go far beyond the SSTRs.
One case that illustrates particularly clearly the different linguistic dimensions
involved in French orthography concerns the relationships between the phoneme
/ s / and the grapheme s, which is its most obvious transcription. The first point to
be underlined is that / s / is one of the most frequent phonemes in French. In the
Véronis (1986) quantitative evaluation, /s/ occupies sixth place in a ranking containing
36 phonemes and represents 5.25% of the total number of occurrences. At the same
time, it is one of the most inconsistent phonemes since it can be spelled as s, c,
ss, f, t and x. In the Véronis (1988) study, the author calculates the "phoneme
transcribability" by SSTR and finds that / s / occupies the fourth poorest place with
72.82% correct transcriptions. Only the phonemes / o / , / a / and / e / presented
poorer scores. Finally, the grapheme s, which is the dominant transcription for
/s/, stands in a rather complex relation to its pronunciation. At the end of words
it serves as a morphosyntactic marker (for nominal plural and as a personal ending
of verbal forms) or is simply part of the lexical definition (for example mais (but)
and mai (the month of May)). In both cases, the final 5 is silent except when it is
followed by a word beginning with a vowel (provided certain syntactic conditions
are met). Word internaly, s stands for /z/ or / s / depending on the context (/z/ when
between two orthographic vowels, otherwise /s/).
The acquisition of the whole system implies the use of complex skills: phonological,
morphological, lexical and syntactic. An adequate experimental design could potentially
allow us to follow the development of each of them, their interrelations and their
failure in the disabled child. The present report is, however, limited to the role played
by the two basic procedures intervening in spelling: lexical and SSTR non-lexical.
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
213
The theoretical framework adopted here retains the most generally agreed aspects
of the spelling models developed during the last 20 years. Most of the models discussed
in the influential book on cognitive processes in spelling edited by U. Frith in 1980
more or less explicitly refer to the two above-mentioned procedures.
It is necessary to add that calling upon these two procedures is not sufficient to
solve the problem of spelling. What is missing from the model is an explicit statement
about the way both procedures interact at different levels. A first level concerns the
development of the whole system. It has been proposed that the orthographic lexicon
develops as a result of the phonological activity (see, e.g. Ehri, 1980, 1991; Gough
& Walsh, 1992; Perfetti, 1991). According to this view, the phonological component
is necessary for the whole system to develop. A second level concerns the contribution
of each process in real-time spelling. For example, as said before, spelling can reflect
morphosyntactic constraints (Carlisle, 1988; Rubin, 1984, 1987; Templeton &
Scarborough-Franks, 1985). The question arises of how these constraints are combined
with the phonological and the lexical procedures, each one sometimes contradicting
the other in the process of spelling words in context.
Another important source of interaction between the lexical and the non-lexical
processes in spelling concerns the involvement of analogies. It has been shown that,
at least in normal spellers, analogical processes intervene in word and in pseudoword
spelling (Campbell, 1983; Barry & Seymour, 1988; Seymour & Dargie, 1990).
Analogical spelling is lexically derived. Therefore, its presence can cast doubt on the
existence of a non-lexical pathway. However, it may be that analogical processes
participate in spelling through the phonological system. This procedure can, indeed,
easily integrate an SSTR system including lexically determined rules»(for example,
the rule /-â^a/ -» -ange can develop under the lexical influence of words like change,
mange, . . .). This account for the effects of analogy assumes that the rules in the
phonological pathway can be lexically activated (e.g. a word like venger will activate
the-*-enge instead of the-» -ange rule). The mere existence of analogical effects in
spelling implies an interaction between the lexical and non-lexical procedures which
needs elaboration.
Finally, one problem that a strict dual-process model cannot easily deal with
concerns the spelling of words involving lexical activity without understanding.
Evidence in ordinary spelling comes from what has been called "slips of the pen"
(e.g. there instead of their) (Ellis, 1979; Hotopf, 1980). According to the classical
dual-process view, the lexical representations of words are activated through their
meaning. In thé example above, the item there has been activated in the lexical store
without any semantic contribution. This may lead to the addition of a third pathway
which allows lexical spelling without meaning. The strongest evidence in favour of
this pathway comes from acquired disorders of spelling (Patterson, 1986).
The literature about acquired spelling disorders has produced strong evidence of
the reality of the two basic spelling procedures discussed so far (see Barry, this volume,
Chapter 2). Many of the "surface" and "phonological" dysgraphics have rather severe
problems with both spelling and reading. More intriguing are the cases of adults
presenting developmental syndromes similar to the acquired syndromes, but having
reached good literacy levels. Campbell and Butterworth (1985) have reported the case
of an undergraduate student with a disorder analogous to the acquired phonological
214
ABNORMAL SPELLING
syndrome (see also Funnell & Davison, 1989, for a similar case). More recently,
Goulandris and Snowling (1991) have described the case of another undergraduate
student presenting a developmental surface dyslexia and dysgraphia syndrome.
Those cases roughly correspond to the taxonomy proposed by Boder (1971): the
"dysphonetic" subjects present similarities with the phonological syndrome and
the "dyseidetic" subjects with the surface syndrome. However, the subjects described
by Boder were less extreme cases than those referred to above.
An important question is whether each of the two basic reading and spelling
procedures can develop independently and afford normal literacy. Bryant and Impey
(1986) have shown that, in a population of normal young readers and spellers, it is
possible to find subjects who are strongly biased towards the use of the phonological
procedure and others who are biased towards the lexical procedure. These subjects
presented a pattern of reading and spelling abilities as extreme as those described
in the literature as being pathological (see, e.g. Temple & Marshall, 1983, for a
developmental phonological case, and Coltheart et al., 1983, for a surface one). They
conclude that each procedure alone can support normal development. This is not
in itself a new idea. The undergraduate students examined by Campbell and
Butterworth and by Goulandris and Snowling demonstrate that it is possible
to compensate a functional deficit in one procedure by an exceptionally good
performance in the other. What is controversial in the Bryant and Impey point of
view is that a substantial proportion of the ordinary school population possess strongly
biased reading and spelling procedures which remain extreme,, suggesting that these
procedures develop independently.
A different model of reading and spelling development supposes that the elaboration
of lexical representations results from the activity of the phonological procedure.
The shift from an alphabetic-phonological reading and spelling procedure to an
orthographic-lexical procedure is a good example of the manner in which the two
basic procedures can be articulated in developmental models. An efficient phonological
procedure is a necessary condition for the development of the lexical procedure. Gough
and Walsh (1992), Freebody and Bryne (1988) and Byrne, Freebody and Gates (1992)
have convincingly argued against the notion that the lexical and the non-lexical
procedures may develop independently to reach levels roughly beyond the third
grade of normal reading and spelling. Exceptional cases (like the subject reported
by Campbell and Butterworth) may be explained by exceptional compensatory
mechanisms.
THE PRESENT STUDY
In order to investigate some of these issues further, we carried out a study to examine
the development of the two basic spelling procedures, the phonological and the lexical
procedures, as well as their relationships in the developmental process, in normal
and disabled children. The study was mainly descriptive and contrasted "consistent"
and "inconsistent" SSTRs. The consistent rules were either contextually independent
(///-» ch independently of contextual factors) or contextually dependent (/g/ followed
by e-* git). In the inconsistent conditions, the target phoneme presented a "dominant"
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
215
transcription (/s/-*s) and a "non-dominant" one (/s/-»c) depending on the
frequency of the "graphoneme". ' For the reasons explained above, the phoneme
/ s / and the grapheme s will be under special focus in the inconsistent condition of
this study (see Alegria & Mousty, in preparation, for further analyses of the /s/-s
association).
The consistency manipulations were aimed at revealing the way in which
phonologically based processes develop. Two extreme conditions were considered.
First, we examined the consistent-context-independent cases, which include SSTRs
that are almost 100% productive. These rules are hypothesized to be adopted
earlier than the inconsistent but dominant rules, which are less productive. Second,
we looked at the inconsistent-non-dominant cases, which inevitably require the
use of lexical procedures. These were assumed to reveal the development of
orthographic representations. In addition, the case of the consistent-contextdependent graphonemes was worth considering, because rules are as productive
as in the consistent-context-independent condition but involve more or less complex
conditions of application. It is plausible that the spelling of those items is not
independent of lexical procedures.
Two groups of subjects were examined: normal and disabled children. The normal
children were second, third and fifth graders of an ordinary school which used a phonic
method of reading instruction. There were 22 subjects in grade 2, 24 in grade 3 and
18 in grade 5. Their mean ages were 7;7, 8;9 and 10;10 years, respectively. The second
graders were tested during the second" part of the school year so that all of them had
reached a spelling level beyond the first steps in phoneme transcription, ;First graders
were not considered because their spelling and reading performances seemed to depend
very much on the particular classroom in which they happened to be (Leybaert et
ai, 1994).
The disabled group came from a special school for children with a specific delay
in literacy acquisition. As is routine procedure in Belgium, these children have been
detected as presenting reading and spelling problems but not intellectual ones (they
must have an IQ of at least 85) by an official institution that decided to send them
to the special school. The school integrated each child in a class corresponding to
his/her reading and spelling level. The subjects came from four consecutive classes that
we shall call A, B, C and D. The number of children per class was 12, 14, 10
and 12, respectively, and their mean ages were 11;3 (10;0-12;6), 11;7 (10;6-12;7),
12;5 (10;5-13;7) and 12;11 (11;5-14;5) years, respectively. These children, even in
class A, had no problems for reading and spelling simple phonological structures.
It is not impossible, then, that a pure phonological dyslexic child would never reach
the classes of disabled children where we began to work.
Subjects were tested collectively in their classroom. They were presented with series
of written sentences with one or more missing words. The experimenter read aloud
the whole sentence and repeated the missing word(s) in turn. The subject's task
was to write down each word at its correct place in the written sentence. The
'According to Véronis (1986), a "graphoneme" is the couple of a grapheme and its phonematic
counterpart. The concept of grapheme corresponds to the "functional spelling units" of Venezky (1970).
So, /%/ -> s (serpent), hi •* c (cigarette) and / s / -* sc (science) are examples of French graphonemes.
216
ABNORMAL SPELLING
words included critical graphonemes that were either consistent or inconsistent. The
consistent items could be context independent or dependent. The inconsistent items
presented a dominant or a non-dominant spelling.
In the consistent-context-independent condition, the following SSTRs were
considered: /v/ -* v; / p / -*p\ / / / ~* ch; A/ -* i; /a/ -» a; /u/ -» on ; and /&/ -» en. These
items allowed comparison of vowel and consonant spelling (/i/, /e/, /a/, /u/ vs
/v/, / p / , / / / ) on the one hand, and simple and complex graphemes (/, a, v, p
vs ou, eu, ch) on the other hand. In addition, the graphonemes (/R/ -* r, /!/ -» /) were
examined in two different positions: they were either included in a consonantic cluster
(e.g. C/R/V) or constituted an isolated consonant at the beginning or end of a syllable
(e.g. /R/V or V / R / ) . This was designed to evaluate the subject's ability to mentally
isolate phonemes before applying SSTRs.
In the consistent-context-dependent condition, the items were: /g/e-i-* gu
(/g/e-i stands for / g / followed by the letters e or /); /k/e-i -» qu;2 and /N/b-p-» m
(/N/ stands for nasal vowels like / 5 / , / a / or /e7 that are spelled with the letter m
before the letters p or b (nombre, chambre, timbre, . . .) and otherwise with the
letter n).
In the inconsistent condition, three cases were considered: (1) #,C/s/e-i (/s/ at
the beginning) of a word or preceded by a consonant) -*s in the dominant version
and c in the non-dominant one; (2) G/z/V (/z/ preceded by an orthographic
vowel noted G) -» s in the dominant case and z in the non-dominant one; and (3)
G/s/e-i -> ss in the dominant case and c in the non-dominant one.
In the analysis of the data, only the critical graphonemes were considered (i.e. if
the word cigarette /sigaRet/ was used to test the /s/-»c rule, only the letter(s)
corresponding to the initial phoneme was taken into account.
Figure 11.1 represents the mean percentage of correct responses for the consistent
items, per condition and per class, in the control and disabled children. The
results strongly depended on the context (see the top graphs). In the contextindependent case, the scores exceeded 90% correct in both groups from the
very beginning. In the context-dependent condition, however, scores were rather
poor. Two ANOVAs, one per group, carried out with context and class as
factors showed that both these effects were highly significant in each group,
as well as their interaction. Further analyses showed that the context effect
was significant for second and third graders but not for fifth graders in the
control group. This effect remained significant, however, even in the most skilled
class of disabled children.
The graphonemes used in the consistent-context-independent condition included
two independent factors: vowel vs consonant and simple (one phoneme-* one letter)
vs complex (one phoneme ~* several letters) cases. The scores as a function of these
variables are presented in the middle graphs. Only the simple vs complex factor was
relevant in the poorest class of the disabled children; performance was better for simple
than for complex items. In all of the other cases, the performance was too high to
show any difference between the conditions.
2
This rule is not entirely consistent: there are some French words with the /k/e-i -» k graphoneme (e.g.
kilogramme, kilomètre, kératine, . . .).
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
217
DISABLED CHILDREN
CONTROL CHILDREN
Context independent
Context dependent
Simple - C
Simple - V
Complex - C
Complex - V
Phoneme
Cluster
Class
Class
1 J.l Percentage of correct responses on consistent items for each class of control
and disabled children.
FIGURE
The bottom graphs show the results for the spelling of the phonemes / R / and
AI (which are among the most consistent French phonemes) when they are included
in a consonantic cluster (e.g. /kR/V) as opposed to when they are not (e.g. /R/V
or V / R / ) . The scores were almost perfect in all groups, except in the less advanced
218
ABNORMAL SPELLING
CONTROL CHILDREN
DISABLED CHILDREN
too
..,^-'
3
§
40
/
J
® /
20-
©
— 0 -
DOM - #,C/s/e.i -> s
- © -
OOM - GITN
-@~
OOM-G/s/e,i->ss
••••©-
NON-DOM - *,CVs/8,l •> o
••••©•••
N O N D O M - G/z/V -> z
--@-
NON-DOM - G/s/e.i -> o
•>
s
40
-20
T "
A
Class
Class
11.2 Percentage of correct responses on mconsistent items for each class of control
and disabled children.
FIGURE
group of disabled children where the cluster condition was slightly, but significantly,
weaker than in the other condition.
The results obtained in the inconsistent condition present a complex pattern
(see Figure 11.2). In two out of the three graphonemes examined, graphoneme I
(ft,C/s/e,i->s/c) and graphoneme 2 (G/z/V-*s/z), the expected dominance effect
clearly appeared, showing that the dominant versions were spelled at a level almost
as good as if they were consistent. The non-dominant version was extremely poor
at the beginning and improved with schooling. The results in the third case
(graphoneme 3: G/s/e,i -» ss/c) were rather poor in both the dominant and the
non-dominant versions from the beginning onwards. Two ANOVAs, one per group,
were performed considering graphoneme (1, 2 and 3), dominance (dominant vs
non-dominant) and class as factors. The critical triple (graphoneme by dominance
by class) interaction was significant in both the control and the disabled group. In
the control group, local analyses revealed a highly significant interaction between
class and dominance for the graphonemes 1 and 2 but not for the graphoneme 3.
In the disabled group, this interaction was significant for the graphonemes 2 and
3 but not for the graphoneme 1. Finally, it is important to note that in the control
group, dominance was highly significant until the third grade but no longer in the
fifth grade. In the disabled group, however, it remained significant even in the most
skilled class. As can be observed in Figure 11.2, this class of disabled children and
the third graders reached equivalent levels of performance. Note that a similar
outcome characterized the results of the consistent conditions (see Figure 11.1).
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
219
This suggests that the spelling level of third graders is reached by the disabled children
with a delay of about 4 years on average.
SUMMARY AND CONCLUSIONS
The present study was confined to the spelling processes that are generally considered
as being the most basic: phonological and lexical. The former are based on a corpus
of rules that allow speech sounds to be translated into graphemes and the latter consist
in a set of word representations containing orthographic specifications that can be
accessed for spelling. Our research has been aimed at examining the way in which
these two systems develop in normal and in disabled French-speaking children. Let
us first consider the conclusions that can be drawn from the results concerning the
first steps into spelling. Afterwards, we shall examine the question of the evolution
of the system.
The First Steps into Spelling Acquisition
The study began with children attending second-grade classes. They had all been taught
to read and spell with an alphabetic method for more than 1 year. They were all
able to spell almost perfectly words using consistent-context-independent SSTRs. This
was also the case even for the less advanced subjects of the disabled group. These
subjects still showed some residual difficulties at spelling the phonemes / R / and / l /
when they were included in consonantic clusters. This might reflect slight phonemic
awareness troubles. The results, however, were rather good (about 90% correct), so
they do not force the conclusion that the less advanced subjects do not possess
metaphonological abilities at a sufficient level. Similarly, it is interesting, in this
context, to recall that the disabled children did not show any difference between vowel
and consonant spelling. Difficulties in the conscious manipulation of phonemes would
probably produce better results for vowels than consonants. It has been shown that
the former are easier to isolate than the latter; this was particularly clear in young
dyslexic children (Morais, Cluytens & Alegria, 1984). Thus, all our subjects possessed
the basic metaphonological abilities that enabled them to analyse utterances into their
segmental constituents and to translate them into graphemes.
Concerning the spelling procedures demonstrated by the less advanced subjects,
the factors examined (consistency, dominance and context) determine the performance
in a complex way but the general pattern of results is rather similar in both groups
of children. A simple account of these results is to suppose that the first steps into
spelling are based on simplified rules. The data reported allow us to make some
statements about the characteristics as well as the possible origin of these rules.
Rules are productivity independent. The evidence comes from two sources. First,
the ccwsw/evjf-context-dependent items produced extremely poor results. Second, some
inconsistent-dominant items (/s/ -*s and /z/~+s) were almost perfect. At the same
time, the corresponding non-dominant cases (/s/ -> c and /z/ -* z) were almost totally
ignored. An interesting exception is the spelling of the intervocalic / s / , whose
dominant version is ss. This item reached about 50% correct responses in the
220
ABNORMAL SPELLING
control group and 30% in the disabled one, far less than the results obtained with
the two other dominant cases, /s/->s and /z/-*s. The typical error was to spell the
intervocalic / s / with j (then pronounced /z/) instead of ss. These results are difficult
to reconcile with an interpretation of the dominance effects in terms of rule frequency
per se: dominant cases are more frequent than non-dominant cases, hence they
are acquired first. It seems more likely that subjects have a default rule like "spell
/ s / and /z/ as 5 everywhere". The other spellings, c and z, which are indeed
non-dominant, but also ss, which is dominant, can be considered as marked cases,
rarely used at the earlier spelling stages we investigated, probably because they have
not yet been taught at school (see below). Then, the default rules involving inconsistent
graphonemes, despite their relatively modest productivity values, are used as if they
were as systematic as the consistent rules, which are almost 100% productive.
Beside the productivity-independent characteristics of the rules, the present study
shows that these rules are also context-independent. The scores on consistent items
were practically perfect with context-independent items (e.g. A/ -* i everywhere) and
almost null if contextual constraints were to be taken into account (i.e. / g / before
e or i~*gu). In these cases, children spelled the items as if it were contextually
independent (i.e. /gi/ -*gi, which is pronounced /3i/).
The notion that young children do use exclusively context-independent SSTRs was
further examined in a pseudoword spelling task including the phonemes / s / and /z/
placed in different contexts (Alegria & Mousty, in preparation). The results for the
disabled children were clear-cut: / s / and / z / were almost always spelled as
5", the default grapheme, whatever the context in which they were included. A control
group of second graders spelled / z / with s, like the disabled ones, but adopted the
grapheme ss to spell / s / in 50% of the cases. More importantly, this score showed
some context sensitivity: ss was slightly preferred to s in pseudoword internal position
while the opposite occurred at the beginning of the pseudowords. Therefore the
tendency to spell / s / with the grapheme .ra, which is marked in a contextually bound
manner, was already present in some particular cases in the less advanced group of
normal subjects.
If it is assumed that the first steps into spelling acquisition discussed so far involve
relying on a set of simplified rules, the question of their origin arises. The notion
that they have evolved through reading and spelling experience seems unlikely. We
have found numerous examples of rules that do not reflect the statistical properties
of French orthography. For example, the less skilled among our subjects systematically
used the rule / / / -» ch, which is almost always correct, but also / s / ~* s, which is not
at all. The most plausible origin of the initial rules is explicit classroom instruction.
Those rules could basically be something like: one phoneme -» one grapheme (e.g.
/$/-*s, /J/^-ch, . . .). They are taught following a programme that reflects the
frequency of both the phonemes and the SSTRs. Finally, initial instruction does not
include contextual constraints. This can easily account for the fact that consistent
and inconsistent but dominant cases produced similar results in spite of their
differences in frequency. Teachers seem to follow a programme that presents the
dominant cases as if they were consistent. As a result, subjects succeeded in spelling
the dominant cases at more than 90% correct but failed almost completely with the
non-dominant cases. Moreover, at the beginning, the teaching programme does not
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
221
consider relatively infrequent segments (/z/ for example), nor contextual constraints,
although they are totally consistent and concern frequent segments. For example, the
n in the grapheme corresponding to all nasal vowels an, en, in, on, ain, etc., changes
into m before b and/>, but this regularity is totally ignored at the beginning. Similarly,
children overgeneralize primitive rules like / g/ -» g and /k/ -* c to contexts where they
are not correct any longer (i.e. /g/e,i->g£-g/), probably because these restrictions
have not yet been taught them. A minor point that could be added concerns the rules
including double graphemes like ch and eu. This factor had no effect on the
performance of second graders but double graphemes still produced slight difficulties
in the disabled group, suggesting that teachers begin with "one segment -» one letter"
rules. Finally, some of the teachers' remarks addressed to us concerning our tests
(e.g. "You can't put this word into the test because it contains something that I have
not yet taught in class") corroborate the existence of a programme.
Referring to a taught system of rules as a psychological explanation of empirical
data can seem rather unsatisfactory. Indeed, it could be purely ad hoc because
programmes are more or less arbitrary, and differ from one another, so that they
could explain the data whatever they are. A less problematic situation has been
investigated by Read (1986) with his young creative spellers, and by Treiman (1993)
with older children attending classes that allow and promote personal involvement
in the discovery of spelling processes. The general conclusion, howevervseems to be
that whatever the path the child adopts or undergoes, spontaneous or imposed from
above, it will initially incorporate an important variability in its detailed composition.
The final product, however, will be a set of rules sufficiently powerful to allow
individual reading and spelling. The elaboration of the mature reading and spelling
system will depend on the personal involvement in those activities. As a consequence,
the teacher's role will rapidly decrease and perhaps disappear. It may be expected
then that the spelling performance reflects the properties of the orthographic system
on the one hand and their interactions with the properties of the cognitive system on
the other. This will be our main concern in the next section. What is perhaps surprising
is that the spelling mechanisms of normal children after more than 1 year of school
seem to reflect nothing more than an explicitly learned set of rules.
Beside the notion that during the first steps into spelling acquisition children do
use naive productivity- and context-independent SSTRs learned at school, we have
not found any evidence of lexically driven spelling procedures. The results observed
with inconsistent-non-dominant items are clear-cut. The performance on these cases,
which need the use of lexical orthographic representations, was indeed extremely poor
in both groups. Further evidence on this matter has been collected in another
experiment in which inconsistent-non-dominant graphonemes were included in words
of high or low frequency (Alegria & Mousty, in preparation). In agreement with the
present interpretation, frequency effects were absent at the lowest reading levels in
the normal as well as in the disabled groups of children.
Subsequent Development of Spelling Procedures
The conditions that were poor at the earlier stages of spelling improved with
schooling in a radically different way in the control and in the disabled groups.
222
ABNORMAL SPELLING
The performance with the inconsistent-non-dominant items increased rather rapidly
between second and third grade in the control group. This "acceleration" was totally
absent in the disabled group. The same difference appeared with the consistent but
contextually determined items. It is important to keep in mind that the present data are
not longitudinal, so that we must be cautious when interpreting the slopes, especially
that of the disabled group. However, the great majority of the disabled children
remained in the special school until the end of primary education, and they moved
from class to class each year with a probability that is, as indicated by the mean ages
of the successive classes, inferior to the control group. So, the slopes translated into
individual developmental tendencies are even smoother than suggested by the figure.
The acceleration in the development of inconsistent-non-dominant items in the
normal group suggests that it is during this period of time that the orthographic lexicon
incorporates a large number of relatively infrequent words. If we assume that the
lexical representations develop as a result of reading activity, the results suggest that
the reading processes have reached a degree of efficiency that allows the control
children to read more extensively.
The mechanism underlying this process is probably self-learning (whose basic
features have been proposed by several authors: Ehri, 1980; Frith, 1980; Gough &
Walsh, 1992; Jorm & Share, 1983; Perfetti, 1991; Stanovich, 1986, among others).
According to this account, phonological processes are the dynamic1 element in the
elaboration of lexical representations and, finally, in the development of the whole
spelling system. All the abilities involved in the elaboration of the system underlying
phonological reading and spelling are necessary for the lexical system to develop,
but they are not sufficient. The disabled children examined in the present study
possessed reasonably good phonological abilities but remained poor at the lexical
level. This suggests that their word identification processes allow them to read but
not to develop an orthographic lexicon efficiently.
A tentative account can be proposed to explain this difference. Studies comparing
reading and spelling abilities in the same children have revealed the existence of a
group of children who are good readers but poor spellers (Bruck & Waters, 1990;
Frith, 1980; Waters, Bruck & Seidenberg, 1985). It has been demonstrated that they
reach high reading levels by using compensatory reading procedures based on linguistic
redundancy (Bruck & Waters, 1990; Waters, Bruck & Seidenberg, 1985). Frith has
argued that these "mixed" subjects identify words using partial instead of full
orthographic cues. The basic feature that characterizes good spellers could be their
word-processing mechanisms, which are based on full cues, in Frith's terms. The
present data are compatible with the notion that word processing based on full cues
is a necessary condition for the elaboration of an orthographic lexicon which is useful
for word spelling. In a subsequent study (Alegria & Mousty, in preparation) designed
to evaluate the lexical spelling procedures more directly, the effect of word frequency
on the spelling of inconsistent-non-dominant graphonemes was examined. The critical
feature of this experiment was that normal and disabled children were matched on
the basis of a reading comprehension test. The results showed stronger frequency
effects in the former than in the latter group. In agreement with the present
speculations, frequency effects which reflect lexically based spelling processes emerge
beyond the abilities that are assessed by a reading comprehension test.
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
223
An account of the difficulties encountered by the disabled group in developing an
orthographic lexicon that can be articulated easily with the previous one has been
recently proposed by Goulandris and Snowling (1991). According to these authors,
surface dyslexia may be due to a non-specific visual memory deficit, The case they
describe is that of a successful reader presenting a developmental surface syndrome
with severe difficulties in tasks requiring the visual recognition of geometric shapes
and the reproduction of Greek letters or abstract letter-like sequences. The authors
assume that this impairment has prevented the subject from developing detailed
orthographic representations of words. This account may in principle explain the use
of partial cues in reading. It simply results from the subject's inability to store full
cues as a consequence of a general visual memory deficit.
To conclude on this point, the disabled subjects examined in the present study have
a profile globally similar to the surface dyslexies. They suffer, as a group, a strong
delay in lexical spelling while their phonological processes are relatively good.
A final point that deserves attention concerns the development of the ability to
spell consistent graphonemes that are contextually constrained. It is striking that these
entirely consistent items showed such a poor performance. The explanation of the
failure at the earlier stages considered was that the corresponding rules had not yet
been taught. Afterwards, the performance may develop as a result of lexical or of
phonological procedures, these now including contextual constraints, or of both.
Consistent-context-dependent items show developmental curves that are almost
exactly the same as those of inconsistent-non-dominant items. This happened in the
normal as well as in the disabled children, even though the evolution of performance
differs from one group to the other. The most economical explanation is that
performance depends exclusively on lexical procedures in both conditions. The lexical
involvement in word spelling when lexical information exists is identical in both
conditions. So, if a second source of information (i.e. the rule-based source) was
involved in the consistent-context-dependent condition, performance should exceed
that observed in the inconsistent-non-dominant condition (where rules are not
involved). The absence of a clear advantage of the former relative to the latter
condition argues in favour of a pure lexical explanation of the results in both
conditions.
This rather surprising conclusion needs some qualification for at least two reasons.
First, this explanation supposes that the frequencies of the words used in both
conditions are equivalent, and this was indeed the case.3 The second reason involves
the way in which the two basic procedures are combined in order to produce spelling.
The explanation just proposed implicitly supposes that the lexical and the rule-based
procedures are hierarchically organized, with the latter intervening solely when the
former does not provide any response. For example, to spell the word guerre (war)
the orthographic lexicon is first consulted. The rule-based procedure will be used only
if an orthographic representation is not found. The response will be correct if the
(correct) orthographic representation is found or if the contextual rule is available
5
The mean logarithmic values of the word frequencies taken from Trésor de la Langue Française (1971)
were 3.17 in the inconsistent-non-dominant condition and 3.62 in the consistent-context-dependent
condition. The difference was non-significant.
224
ABNORMAL SPELLING
(/g/e -» gu). In the other cases, the response will be an error (produced, for example,
by the use of the primitive context-independent rule / g / -» g). According to this model,
the effects of rules are simply added to those of the lexical activity. The tentative
conclusion concerning the consistent-context-dependent condition, stating that
"nothing is added", depends on the adhesion to the notion that the orthographic
lexicon is consulted first, before the rules, which are applied in turn, from the more
constrained to the less constrained.
More complex modes of interaction between the lexical and non-lexical procedures
can be conceived that are likely as the one just discussed. For example, the availability
of rules as well as of orthographic representations may depend on their respective
frequency, so that some rules could "overtake" the orthographic representation and
impose their spelling. In the case of the word guerre, the naïve / g / -» g rule could
be so strong that it eventually hinders the activation of both the orthographic word
representation and the contextual rule. The results obtained in the consistent-contextdependent condition may be exceptionally poor because the rules examined have more
primitive competitors that are especially strong. The evolution of spelling could depend
on the inhibition of these rules. The questions raised are important because they
concern one of the less understood problems: that of the interactions between the
two basic spelling procedures. The experiments currently undertaken are aimed at
collecting data to constrain speculation on this matter.
ACKNOWLEDGEMENTS
The study reported in this chapter was supported by the National Fund for Scientific Research
("Loterie Nationale", convention 8.4505.92) and by the Ministry of Education of the Belgian
French-speaking Community (Concerted Research Action, convention 91/96-148), We are
grateful to the staff and children from the schools "Etablissement d'Enseignement Primaire
Spécial" (Court-St-Etienne) and "Ecole de la Sainte Famille" (Braine-Lalleud).
REFERENCES
Alegria, J. & Mousty, P. (in preparation). The development of spelling procedures in
French-speaking, normal and disabled children: effects of frequency and lexicality.
Barry, C. & Seymour, P. H. K. (1988). Lexical priming and sound-to-spelling contingency
effects in nonword spelling. Quarterly Journal of Experimental Psychology, 40A, 5-40.
Boder, E. (1971). Developmental dyslexia: a diagnostic screening procedure based on three
characteristic patterns of reading arid spelling. In B. Bateman (Ed.), Learning Disorders,
Vol. 4. Seattle: Special Child Publications.
Bruck, M. & Waters, G. S. (1990). An analysis of the component spelling and reading skills
of good readers-good spellers, good readers-poor spellers, and poor readers-poor spellers.
In T. H. Carr & B. A. Levy (Eds), Reading and its Development. San Diego, CA: Academic
Press, pp. 161-206.
Bryant, P. & Impey, L. (1986). The similarities between normal readers and developmental
and acquired dyslexies. Cognition, 24, 121-137.
Byrne, B., Freebody, P. & Gates, A. (1992). Longitudinal data on the relations of word-reading
strategies to comprehension, reading time, and phonemic awareness. Reading Research
Quarterly, 27, 141-151,
SPELLING PROCESSES IN FRENCH-SPEAKING CHILDREN
225
Campbell, R. (1983), Writing nonwords to dictation. Brain and Language, 19, 153-178.
Campbell, R. & Butterworth, B. (1985). Phonological dyslexia and dysgraphia in a highly literate
subject; a developmental case with associated deficits of phonemic awareness and processing.
Quarterly Journal of Experimental Psychology, 37A, 435-475.
Carlisle, J. F. (1988). Knowledge of derivational morphology and spelling ability in fourth,
sixth, and eighth graders. Applied Psycholinguistics, 9, 247-266.
Catach, N. (1980). L'orthographe Française: Traité Théorique et Pratique. Paris: Nathan.
Coltheart, M., Masterson, J., Byng, S., Prior, M. & Riddoch, J. (1983). Surface dyslexia.
Quarterly Journal of Experimental Psychology, 35A, 469-495.
Ehri, L. C. (1980). The development of orthographic images. In U. Frith (Ed.), Cognitive
Processes in Spelling. London: Academic Press.
Ehri, L. C. (1991). Learning to read and spell words. In L. Rieben & C. A. Perfetti (Eds),
Learning to Read. Hillsdale NJ: LEA, pp. 57-63.
Ellis, A. W. (1979). Slips of the pen. Visible Language, 13, 265-282.
Freebody, P. & Byrne, B. (1988). Word-reading strategies in elementary school children:
relations to comprehension, reading time, and phonemic awareness. Reading Research
Quarterly, 23, 441-453.
Frith, U. (1980). Unexpected spelling problems. In U. Frith (Ed.), Cognitive Processes in
Spelling. London: Academic Press.
Funnel!, E. & Davison, M. (1989). Lexical capture: a developmental disorder of reading and
spelling. Quarterly Journal of Experimental Psychology, 41A, 471-487.
Gak, V. G. (1976). L'orthographe du Français, Essai de Description Théorique et Pratique.
Paris: Selaf.
Gleitman, L. R. & Rozin, P; (1977). The structure and acquisition of reading I: relations between
orthographies and the structure of language. In A. S. Reber & D. L, Scarborough (Eds),
Toward a Psychology of Reading: The Proceedings of the CUNY Conference. Hillsdale,
NJ: LEA.
Gough, P. B. & Walsh, M. A. (1992). Chinese, Phoenicians and the orthographie cipher of
English. In S. Brady & D. Shankweiler (Eds), Phonological Processes in Literacy. Hillsdale,
NJ: LEA, pp. 199-210,
Goulandris, N. K. & Snowling, M. (1991). Visual memory deficits: a plausible cause of
developmental dyslexia? Evidence from a single case study. Cognitive Neuropsychology,
8, 127-154.
Hotopf, W. H. N. (1980). Slips of the pen. In U. Frith (Ed.), Cognitive Processes in Spelling.
London: Academic Press.
Jorm, A. F. & Share, D. L. (1983). Phonological recording and reading acquisition. Applied
Psycholinguistics, 4, 103-147.
Klima, E. S. (1972). How alphabets might reflect language. In J. F. Kananagh & I. G. Mattingly
(Eds), Language by Ear and by Eye. Cambridge, Mass: MIT Press.
Leybaert, J., Alegria, J., Deltour, J. J. & Skinkel, R. (1994). Apprendre à lire. Rôle du langage,
de la conscience métaphonologique et de l'école. In J. Grégoire & B. Piérart (Eds), Évaluer
les troubles de la lecture: Les nouveaux modèles théoriques et leurs implications diagnostiques.
Bruxelles: De Boeck.
Morais, J., Cluytens, M. & Alegria, J. (1984). Segmentation abilities of dyslexies and normal
readers. Perceptual and Motor Skills, 58, 221-222.
Patterson, K. (1986). Lexical but nonsemantic spelling? Cognitive Neuropsychology, 3,
341-367.
Perfetti, C. A. (1991). Representations and awareness in the acquisition of reading competence.
In L. Rieben & C. A, Perfetti (Eds), Learning to Read. Hillsdale, NJ: LEA, pp. 33-44.
Read, C. (1971). Pre-school children's knowledge of English phonology. Harvard Educational
Review, 41, 1-34.
Read, C, (1986). Children's Creative Spelling. London: Routledge & Kegan Paul.
Rubin, H. (1984). An investigation of (he development of morphological knowledge and its
relationship to early spelling ability. Unpublished doctoral dissertation, University of
Connecticut, Storrs, CT.
226
ABNORMAL SPELLING
Rubin, H. (1987). The development of morphological knowledge in relation to early spelling
ability. Haskins Laboratories Status Report on Speech Research, SR-89/90, 121-131.
Seymour, P. H. K. & Dargie, A, (1990), Associative priming and orthographic choice in
nonword spelling. The European Journal of Cognitive Psychology, 2, 395-410.
Stanovich, K. E. (1986). Matthew effects in reading: some consequences of individual differences
in the acquisition of literacy. Reading Research Quarterly, 21, 360-407.
Temple, C. M. & Marshall, J. C. (1983). A case study of phonological dyslexia. British Journal
of Psychology, 74, 517-535.
Templeton, S. & Scarborough-Franks, L. (1985). The spelling's the thing: knowledge of
derivational morphology in orthography and phonology among older students. Applied
Psycholinguistics, 6, 371-390.
Treiman, R. (1993), Beginning to Spell: A Study of First-Grade Children. New York: Oxford
University Press.
Trésor de la Langue Française (1971). Paris: Klincksieck.
Venezky, R. L. (1970). The Structure of English Orthography. The Hague: Mouton.
Véronis, J. (1986). Etude quantitative sur le système graphique et phono-graphique du français.
Cahiers de Psychologie Cognitive, 6, 501-531.
Véronis, J. (1988). From sound to spelling in French: simulation on a computer. Cahiers de
Psychologie Cognitive, 8, 315-334.
Waters, G. S., Bruck, M. & Seidenberg, M. (1985). Do children use similar processes to read
and spell words? Journal of Experimental Child Psychology, 39, 511-430.